Participants Dept. of Mathematical Sciences, Aalborg University: E.Susanne Christensen, Susanne G. Bøttcher Dept. of Forensic Genetics, University of Copenhagen: Niels Morling, Helle Smidt Mogensen Dept. of Statistics, Oxford University: Steffen L. Lauritzen Abstract This project investigates the behavior of the PCR Amplification Kit. A number of known DNA-profiles are mixed two by two in "known” proportions and analyzed. Short Tandem Repeats (STR) Human identity tests focus on Short Tandem Repeat markers (STR markers). STR markers are genetic loci consisting of repeated subunits, 2-8 base pairs in length. Discrimination between individuals is possible because the number of subunits present for a given marker varies from person to person. Simultaneous analysis of several STR markers allows for the compilation of a profile, which is almost unique to a given individual. Gamma distribution models are fitted to the resulting data to learn to what extent actual mixing proportions can be rediscovered in the amplifier output and thereby the question of confidence in separate DNA profiles suggested by an output is addressed. Weight of DNA- British data vWA 1 -1 0 1 0 -1 std residual DNA mixture output std residual 2 2 FGA 1.0 1.5 0.5 1.0 weight weight D3 D18 1.5 0.53 x 10ˉ3 3.38 x 10ˉ4 2.95 x 10ˉ4 2.53 x 10ˉ4 1.80 x 10ˉ5 2.90 x 10ˉ4 2.50 x 10ˉ4 7.13 x 10ˉ5 5.42 x 10ˉ4 1.70 x 10ˉ5 0.5 1.0 2 1 1.5 0.5 weight Amount of DNA- Danish data FGA vWA 0 0 DNA amount DNA amount D3 D18 Er Ly Va Ca Er Ly Va Er Ly Va Std. residuals 4 2 Bi Er Ly Va -2 -2 0 Std. residuals 1.0 0.5 0.0 Bi 4 6 Mix with Ca, vWA 50 30 10 -0.4 -10 0 -10 Ca 100 200 300 400 500 2 Mix with Ca, FGA 100 200 300 400 500 0 Mix with Bi, vWA 0 100 200 300 400 500 0 100 DNA amount Mix with Bi, D18 500 DNA amount 6 Discussion Er Ly Va 4 2 0 0 -2 -1 -2 Ca Ca Er Ly Va Bi Mix with Er, vWA Er Ly Va Va Ca Ly Va 1.0 Ca Er Va Ca Er Va 4 Mix with Ly, D18 3 4 4 Bi Mix with Ly, D3 6 1 0 -1 0 2 2 2 2 4 Mix with Ly, vWA 0.0 Bi Mix with Er, D18 6 Mix with Er, D3 Va 0 Bi 6 Ly Ly 0.5 40 20 0.5 0.0 Ca Er 60 1.0 60 40 20 0 Bi Bi Mix with Ly, FGA 80 80 Mix with Er, FGA -2 0 0 Data were created by an controlled laboratory experiment performed by Section of Forensic Genetics, University of Copenhagen to investigate the performance of the AmpF1STRSGM Plus PCR Amplification Kit (Applied Biosystems, CA, USA) in STRprofiling. Samples were created from 4 persons with known profiles. Mixtures of two contributors with different but known amounts of DNA were created as a full factor experiment. Also one-contributor samples were analyzed for all four persons in different concentrations. Every sample were analyzed twice. Observations used from this dataset are peak height, which are approximately proportional to peak area in this data. -2 The Cph-Crime-SGMP-Mix-Exp-2005-1 dataset -1 0 1 2 2 0 200 300 400 Mix with Ca, D18 6 Mix with Ca, D3 3 4 Mix with Bi, D3 1.5 weight 6 Mix with Bi, FGA 1.0 1.0 1.43 x 10ˉ3 1.20 x 10ˉ3 1.09 x 10ˉ3 7.73 x 10ˉ4 1.80 x 10ˉ5 1.29 x 10ˉ3 1.13 x 10ˉ3 4.69 x 10ˉ5 1.24 x 10ˉ3 2.40 x 10ˉ5 0.5 x 10ˉ2 x 10ˉ3 x 10ˉ3 x 10ˉ3 x 10ˉ4 x 10ˉ3 x 10ˉ3 x 10ˉ3 x 10ˉ2 x 10ˉ4 0.0 1.04 6.90 5.93 4.92 1.90 6.39 5.70 5.48 1.20 2.06 Std. residuals 10ˉ2 10ˉ2 10ˉ2 10ˉ2 10ˉ3 10ˉ2 10ˉ1 10ˉ2 10ˉ2 10ˉ4 80 x x x x x x x x x x 60 2.88 2.53 2.26 1.57 0.53 2.46 0.26 2.29 2.84 6.93 40 7.99 x 10ˉ6 6.33 x 10ˉ6 2.38 x 10ˉ5 1.44 x 10ˉ5 1.18 x 10ˉ5 8.35 x 10ˉ6 5.03 x 10ˉ6 6.99 x 10ˉ6 1.24 x 10ˉ5 5.16 x 10ˉ6 Std. residuals- Danish data 1 The constructed two-persons mixtures divides the persons in two groups, forming the mixtures: G.J and C.J, (involving three persons), and C.A, S.A, H.A and C.P, (involving four other persons). Observations in this British data are peak area. x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 x 10ˉ1 std .err ( ˆ ) 20 1.60 2.35 4.67 2.26 3.29 2.20 1.82 1.93 2.03 2.28 std.err (ˆ ) 0 10ˉ5 10ˉ5 10ˉ6 10ˉ5 10ˉ4 10ˉ5 10ˉ5 10ˉ5 10ˉ5 10ˉ5 ˆ ̂ Std. residuals x x x x x x x x x x 10 Data have been processed by Genescan™ software, (Applied Biosystems). Systems analyzed are: D3S1358, vWA, D16S539, D2S1338, AMEL, D8S1179, D21S11, D18S51, D19S433,THO1,FGA. 5.55 4.30 2.05 5.63 1.02 6.00 4.30 5.05 9.01 3.29 system THO D16 D18 D19 D2 D21 D3 D8 FGA vWA 4 Mixtures have been created from two contributors with known DNA profile, known mixture rates and known concentrations at 0.5 ng/µl, and seven different mixing proportions. 1.44 1.90 4.50 1.88 3.26 1.91 2.06 1.75 1.83 1.83 std .err ( ˆ ) std.err (ˆ ) ˆ 0.0 0.2 0.4 The QLB dataset ̂ 20 Source: Forensic DNA typing. J. M. Butler system THO D16 D18 D19 D2 D21 D3 D8 FGA vWA -2 -1 -1 0 1 std residual Estimates- Danish data Estimates- British data 0 std residual 2 3 3 4 0.5 Bi Ca Ly Va Bi Ca Ly Bi Va Ca Er Va Bi Ca Er Va Std. residuals- British data Notation m: system, e.g. FGA i : person j : sample a : type of allel nm( j): number of allel type a (0, 1 or 2) ia wm( j): peak height or peak area ia ( j): "contribution" of DNA i m( j) wm ( j) wia i a kam( j) ( j)nm( j) ia i i Residuals reveals the need for incorporation of a personal effect in the model, beyond what is included in amount of DNA present. The additional effect is seen in both onecontributor samples and in mixtures. There is no obvious difference in using mixing proportions or actual amount of DNA in the parameterization. In both cases it is seen, that some moderation from proportionality should be considered as residuals decrease with increasing amount of DNA in the sample. References L( W | , n , , ) ( n , ), ia i ia i ia 1 w f (w) w e ( ) E(W | , na, , ) n ia i ia V(W | , na, , ) n 2 E(W | , na, , ) ia i ia ia This ensures that: i ( j)nia a i j ˆ ˆ a j wa ( j) J. Mortera, A.P.David, S.L. Lauritzen, Probabilistic expert systems for DNA mixture profiling, Theor. Popul. Biol. 63 (2003), 1919-205 M.W. Perlin, B. Szabady, Linear mixture analysis, a mathematical approach to resolving mixed DNA samples, J. Forensic Sci. 45 (2001), 1372-1378 The parameters are estimated by solving the following maximum likelihood equations, B.S. Weir, C.M. Triggs, L.Starling, L.I. Stowell, K.A.J. Walsh, J.S. Buckleton, Interpreting DNA mixtures, J.Forensic Sci. 42(5), (1999), 987-995 R.G Cowell, S.L.Lauritzen, J.Mortera, Identification and separation of DNA mixtures using peak area information, Forensic Sci. Int. To appear. Estimation i ( j)nia ˆ (i ( j)nia ) i ( j)nia (ln wa ln ˆ ), a j i a j i i where is the digamma function. From the Danish data it is clear that a full understanding and modeling of the STR- amplification still is far from being obtained. It is however of great importance that this work is carried out in order to address the question of separation of DNA profiles from mixture samples. J.M Butler, Forensic DNA typing. Elsevier, USA. 2005 with mean and variance given as: The findings in Cph-Crime-SGMP-Mix-Exp-2005-1 data are not quite as consistent, but there is a clear tendency of over- or underestimation throughout the systems for any two persons, as there are systematic differences between individuals analyzed. The tendency to decreasing std. residuals for increase in amount in DNA is found in the Danish data as well. The model seems logical and produces a fair fit to data. The model is defined as follows: L(Wa | , na, , ) n , i i ia In the QLB-data mixture C.J was consistently overestimated, S.A was overestimated in 9 of 10 markers, H.A was underestimated for 6 markers and mainly so for 3 others, and the C.P mixture was consistently underestimated for 2 markers and mainly underestimated for 3 markers. There is a slight tendency to decreasing std. residuals for increasing weight of DNA in the sample. Conclusion Model The plots shows results from four systems only. T. Wang, N. Xue, R. Wickenheiser, Least square deconvolution (LSD); a new way of resolving STR/DNA mixture samples, in: Proceedings of the 13’th International Symposium on Human Identification, October 7-10, Phoenix, AZ, 2002. I. Evett, P.Gill, J. Lambert, Taking account of peak ares when interpreting mixed DNA profiles, J. Forensic Sci, 43 (1998), 62-69 Future works A parameter of ’personal impact’ shall be incorporated in the model. The important question of ’drop-outs’, ’drop-ins’ and stutters shall be considered. Estimation of combined ’personal impact and amount DNA’ shall be considered for the purpose of separation of DNA – profiles in mixture DNA samples from crime scenes on basis of the model. Sensibility to the actual machine should be investigated, and influence of injection times be considered. A calibration algorithm for adaption to machinery will be constructed. Contact: E. Susanne Christensen, e-mail: susanne@math.aau.dk